In this paper we have presented the most recently available performance data on a new supercomputer which is currently under development at ACRI. Our main aim in so doing has been to highlight the execution efficiency that derives from the high level of decoupling in this novel architecture. The measurements indicate that the actual frequency of LOD events, as generated by the pre-production compiler, leads to execution time penalties that for most programs represent a small fraction of the minimum possible execution time. We see this as a vindication of the principle of decoupling embodied in this architecture, and as a foretaste of the potential that decoupling provides for latency hiding in high performance computers.